How to aggregate by date when a full timestamp is given in aggregation framework?

Question

I have a collection of errors, so that every error carries a date field. How can I aggregate/count/group the errors by DAY only (i.e. exclude the time of the da

Answer 1

You can convert your timestamp field to string with specific date format by using $project (aggregation)

aggregate([
    {
        '$project': {
            newFieldName: {'$dateToString': {format: '%Y-%m-%d', date: '$yourDateFieldName'}}
        }
    }, {
        '$group': {
            _id: {newFieldName: '$newFieldName'},
            viewCount: {'$sum': 1}
        }
    }, 
    {'$sort': {'_id.newFieldName': 1}}
], {})

Answer 2

You can do this by using the following aggregation operators:

• $group
• $year
• $month
• $dayOfMonth

This gives you the error count for each date:

db.errors.aggregate(
    { $group : {
        _id: {
            year : { $year : "$date" },        
            month : { $month : "$date" },        
            day : { $dayOfMonth : "$date" },
        },
        count: { $sum: 1 }
    }}
);

This example assumes that the date field in your error documents is date and of type BSON Date. There is also a Timestamp type in MongoDB, but use of this type is explicitely discouraged by the documentation:

Note: The BSON Timestamp type is for internal MongoDB use. For most cases, in application development, you will want to use the BSON date type. See Date for more information.

Answer 3

Since mongo 2.6 is there $dateToString function that can be used in aggregate like that:

db.errors.aggregate([   
 {
   $group : {
    _id: { $dateToString: { format: "%Y-%m-%d", date: "$date" }},
    count: { $sum: 1 }
   }
 }
]);